Ublažavanje utjecaja zlonamjernih čvorova u mrežama raspodijeljenih glavnih knjiga s čvorovima ograničenih resursa

Abstract

Distributed Ledger Technology (DLT) allows the maintenance of a continuously growing ledger in a peer-to-peer network. Nodes participating in the network require significant computational power and memory to download and verify the ledger. To solve this problem, light clients were developed to verify the integrity of the ledger by downloading only a subset of the ledger, but they are vulnerable to malicious influence. The thesis presents Aurora, a set of three stochastic algorithms that add trustlessness to DLT networks by analyzing the network structure and discovering sets that contain honest nodes that can be used for future interactions. The solution allows a new node to join the network and initiate ledger download with an honest node, or to check the presence of a transaction without downloading the entire ledger or a subset of the ledger. The work provides a comprehensive overview of DLT, pseudocode for the solution, and an evaluation procedure using open-source DLT simulation tools. Furthermore, the solution was integrated into Trinity, a Python-based DLT client for the Ethereum network, and was evaluated on the Ethereum production network, consuming approximately 0.31 MB of RAM and 1 MB of storage at runtime. The proposed solution outperforms other state-of-the-art solutions and can be deployed on resource-constrained devices. It incentivizes users to interact with the ledger in a trustless manner. Overview of scientific contribution: 1. A new probabilistic honest set creation algorithm 2. A new probabilistic transaction history synchronization algorithm 3. A new probabilistic transaction presence checking algorithm 4. A new evaluation procedure in a resource constrained environment